[monodroid] Prevent a segfault when Java class name is unavailable #5623
Conversation
|
I think another possibility (4.) is that the |
Context: dotnet#5619 It appears that Xamarin.Android application uploaded to Google Console internal Test Track can fail the pre-launch test because the `MonodroidRuntime::get_java_class_name_for_TypeManager` method, can sometimes get a `nullptr` Java class name from JNI: Build fingerprint: 'google/blueline/blueline:9/PQ3A.190801.002/5670241:user/release-keys' Revision: 'MP1.0' ABI: 'arm64' pid: 15773, tid: 15773, name: com.Myapp >>> com.Myapp<<< signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0x0 Cause: null pointer dereference x0 0000000000000000 x1 0000000000000000 x2 0000000000000000 x3 00000000000000e1 x4 0000000000000000 x5 0000000000000305 x6 0000000000000006 x7 0000000000000010 x8 0101010101010101 x9 4adfd1b7ed2e23e2 x10 0000000000430000 x11 000000000000000a x12 0000000000000002 x13 0000000000000001 x14 00000000000000a8 x15 000000000000000a x16 0000007f703f00f8 x17 0000007f7031f290 x18 0000000000000008 x19 0000000000000000 x20 0000007eecae0460 x21 0000000000000000 x22 0000000000000000 x23 0000000000000099 x24 0000000000000085 x25 0000007eecb30010 x26 0000000000000099 x27 0000000000000000 x28 0000007f724255e0 x29 0000007fe01ad010 sp 0000007fe01acff0 lr 0000007f7036f3b4 pc 0000007f7031f2a0 backtrace: #00 pc 000000000001e2a0 /system/lib64/libc.so (strlen+16) #1 pc 000000000006e3b0 /system/lib64/libc.so (strdup+20) #2 pc 000000000000d4e8 /data/app/com.Myapp-1lufsJe3FIZ0YHoEAB08KA==/split_config.arm64_v8a.apk (offset 0x165d000) (xamarin::android::internal::MonodroidRuntime::get_java_class_name_for_TypeManager(_jclass*)+92) #3 pc 000000000005f6d0 <anonymous:0000007ecc4ac000> This can happen if either one of the below points is true: 1. The Java [`Class.getName`](https://developer.android.com/reference/java/lang/Class?hl=en#getName()) method is absent 2. `Class.getName` returns `nullptr` (a nameless class?) 3. `env->GetStringUTFChars` returns `nullptr` (a memory allocation failure?) Out of these, 3. appears to be most probable and this commit adds a check for a `nullptr` pointer there, failing gracefully instead of segfaulting. This is NOT a fix for the original problem as we don't know by what it is caused but, nevertheless, the `nullptr` check should be there.
grendello
force-pushed
the
issue-5619
branch
from
89d8fd6 to
57051a5
Compare
if However, I updated the code to at least print an error when |
3 participants
Context: #5619
It appears that Xamarin.Android application uploaded to Google Console
internal Test Track can fail the pre-launch test because the
MonodroidRuntime::get_java_class_name_for_TypeManagermethod, cansometimes get a
nullptrJava class name from JNI:This can happen if either one of the below points is true:
Class.getNamemethod is absent
Class.getNamereturnsnullptr(a nameless class?)env->GetStringUTFCharsreturnsnullptr(a memory allocationfailure?)
Out of these, 3. appears to be most probable and this commit adds a
check for a
nullptrpointer there, failing gracefully instead ofsegfaulting. This is NOT a fix for the original problem as we don't
know by what it is caused but, nevertheless, the
nullptrcheck shouldbe there.